Log file Viewer
#mongodb logs for Wednesday the 19th of February, 2014
[02:25:05] <RaceCondition> and query whose resultset should be a subset of the first one, is returning 1 document, which is odd
[02:27:03] <ruphos> your $or syntax is incorrect: http://docs.mongodb.org/manual/reference/operator/query/or/
[02:28:48] <RaceCondition> Champi: ah... OK; this query was generated by Lift-Rogue not myself... so it's a bug in the library
[02:36:06] <murkk> Anyone able to answer what I suspect to be an easy pymongo question around finding entries?
[02:36:49] <murkk> How does db.entries.find( { "expires": { $gt:"2014-02-10T15:10:00-06:00"}}) look in python?
[02:39:08] <murkk> entries = mongo.db.entries.find({"expires": { $gt:"2014-02-10T15:10:00-06:00"}}) is definitely not it
[03:04:14] <cheeser> i would think you'd need to create a date type rather than that string for a date comparison
[03:20:06] <murkk> I actually am on the python side: entries = mongo.db.entries.find({"cap:expires": { $lt: datetime.now()}}) and get a syntax error at $lt
[03:38:50] <fourq_> Which property of var db = mongoose.createConnection(...); has the native Mongodb connection?
[04:36:08] <the8thbit> Does mongo distinguish between database structure and the actual information in the database?
[04:36:39] <the8thbit> It would be nice to have the structure of the database on github, without any of the elements in that structure
[04:40:02] <the8thbit> I think it would be nice to have the schema on github, but none of the content there, so that e.g., email addresses aren't shared publically
[04:40:51] <cheeser> edsiper: i'm not a C guy but i thought there was one. or maybe that it was under development.
[04:42:59] <the8thbit> cheeser: ha, yeah, the goal isn't social networking, but its a very social-networkign-like app. User accounts, private messages, etc...
[05:43:36] <murkk> cheeser: I believe quotes around the $lt did it, well the syntax error went away.
[10:26:28] <AlexejK> Hi, got a question regarding MongoDB-Java driver.. We have an app that is two different WAR files in one same app-server. Both WAR files create a MongoClient instances and it works fine in 2.11.4 version of the driver but with the 2.12.0-rc0 it seems to be throwing warnings" Unable to register MBean org.mongodb.driver:type=ConnectionPool,clusterId=3,host=localhost,port=27017. javax.management.InstanceAlreadyExistsExce
[10:26:28] <AlexejK> ption". I know this is an RC0 but trying to figure out what has changed and what we can do to avoid this, if this is an intended change int he driver
[11:29:34] <MmikePoso> Hi, guys. Has anyone been able to capture mongo traffic and then replay it against another system using mongosniff?
[12:10:31] <adimania> hi. I have a standalone mongodb. I want to move it to replica set and add a slave node to it. I want to know that while the data propagates to the slave, will the mongo master be usable?
[12:28:54] <dmarkey> Anyway, does this aggregation make sense.. product_analytics.aggregate([ { '$group': { "_id" : "$metrics", 'average_logon_times' : { "$avg" : "metrics.logon_times" } } } ] )
[12:29:40] <dmarkey> $metrics being a list of dicts, that contains a logon_times key (which is just an int)
[12:39:31] <kali> dmarkey: it's a point of irc etiquette really. there are hundreds of people on the channel, you can reasonably expect that there is somebody who can answer questions about aggregation. on the other hand, asking that somebody will step up and present himself as an "expert" is bound to fail...
[12:42:24] <Derick> do you want the average over all the items in all metrics lists in all documents?
[12:45:33] <joannac> But yes, unwind the array, then do your aggregation (assuming you want a global average)
[12:49:20] <dmarkey> aggregate([ {'$unwind' : "$metrics" } , { '$group': { "_id" : "$metrics", 'average_logon_times' : { "$avg" : "logon_times" } } } ] )
[12:53:31] <kali> dmarkey: what about: aggregate([ {'$unwind' : "$metrics" } , { '$group': { "_id" : "all", 'average_logon_times' : { "$avg" : "metrics.logon_times" } } } ] )
[12:58:20] <d-snp> I have a blurb of text that gets inserted into the database a lot of times, so I want to deduplicate it, my current idea is to calculate a hash for it on the app side, and then do an upsert into the database, with the hash as a key
[12:58:34] <d-snp> is that a good idea, or should I use some mongodb built in thing to calculate the hash of the text?
[13:01:00] <d-snp> okey, so the hash function I'm using is ruby's String#hash, which is rather fast but I'm not sure if it returns unique enough keys
[13:04:08] <kali> d-snp: i would use md5. i would not trust ruby implementation to provide consistent hash values across ruby versions bump
[13:04:34] <kali> d-snp: also, if at some point you need to work on your DB outside of the ruby world, ...
[14:02:09] <dmarkey> kali: :( {u'ok': 1.0, u'result': [{u'_id': u'all', u'average_logon_times': 0.0}]}
[14:03:47] <kali> dmarkey: can you dump your document in the shell instead of python, so that I can try ?
[14:21:23] <Nodex> just an fyi if you want a prettier print from a find() just add .pretty() .. db.foo.find().pretty()
[14:23:43] <d-snp> weird, did you guys know that if you have an _id that's not a valid ascii/whatever string a collection.update fails?
[14:24:53] <d-snp> I accidentily passed a raw md5 digest, and it didn't like it at all, perhaps it's a ruby driver thing
[14:43:33] <kali> d-snp: or use the string representation. i think Binary is not compatible with the aggregation framework, so maybe sticking to the string representation is a better idea
[14:56:55] <alkar> The MMS agent reports strange numbers of memory (way too much than what is guarranteed) for my mongod replica running in OpenVZ (latest version so it /should/ be safe). Is that something I have to worry about - is there anything that my VPS provider can configure for me to fix this?
[15:29:36] <remonvv> I'm curious; anyone any opinions on this : https://jira.mongodb.org/browse/SERVER-12694 . Current MongoDB approach seems broken.
[15:30:16] <kali> dmarkey: not per se, but it can lead to memory explosive pipeline if it is follow by a non streamable operator (like a sort or a non trivial group)
[15:50:09] <movedx> We have a 2.4.6 cluster. .6 has a few security issues, so we want to upgrade to .9. What's the best way of upgrading to .9 across a cluster? Should I just shutdown the router process and then everything on the other boxes and upgrade?
[15:50:24] <movedx> I assume taking the router process down gracefully will precent data loss and future read/writes?
[16:10:01] <alkar> @Number6: currently the databases are empty, yet the secondary memory stats (res,virt,mapped) are much higher than the one in primary and are actually
[16:16:34] <alkar> basically I dont understand why the numbers are so different between the primary and the secondary
[16:30:48] <blizzow> I have two mongos on the same instance. They are hidden members of replica set 0 and replica set 1. mongodb for rs0 catches up(slowly) and suffers virtually no slave lag. mongodb for rs1 consistently falls behind. Yesterday, I started the hidden member of rs1 and watched it catch up. I come in this morning and the member is 1400 seconds behind. I even tried stopping the other hidden members and it still does not catch up. I've e
[17:34:28] <Number6> alkar: It should, but OpenVZ does funky stuff with memory, so having the numbers different between your non-OpenVZ and OpenVZ nodes is expected
[17:35:15] <Number6> I really really dislike that hypervisor because of it's memory management - although it's been a while since I've used it personally
[17:35:19] <alkar> @Number6: I suspect so, however mapped memory (if I understand it correctly) is the size of virtual address space that mongod uses to map the disk data - as such it should be consistent between the replicas
[17:36:22] <Number6> alkar: It should. Since MongoDB works this out using Linux system calls, it relies on the VM to correctly report memory usage
[17:36:23] <alkar> @Number6: I do as well. There's one thing it's definitely good at: overselling :) but that was my low budget solution for a replica set
[17:36:59] <Number6> alkar: If you notice between your openVZ hosts, the memory looks the same on both - so my money is on a reporting issue
[17:37:11] <alkar> @Number6: so I guess the only way to tell is migrate the data and run some tests - see how they behave
[17:37:55] <alkar> @Number6: btw, is there a way to view all 3 hosts on the same page when viewing the replica set graphs? (3 column thing instead of 2)
[17:45:32] <alkar> @Number6: unless I click on the third replica which then shows me that one by itself
[17:46:00] <Number6> alkar: Ok, try "Dashboards" >> "test". I did you up a really rough and ready one
[17:46:48] <Number6> Dashboards can be customsied under "Dashboards" >> "add chart" >> Selecting the desired options >> Give it a name >> Save
[17:55:01] <alkar> @Number6: so dashboard is the way to go cool tyvm! (delete yours by mistake - I thought wtf is "test" hehe)
[17:55:23] <alkar> @Number6: all's good mate cheers for the help in this, it probably has to do with the hypervisor
[17:55:55] <foobarius> Trying to install latest perl modules (0.702.2) on ubuntu 12.04 system. Test t/db.t line 52 failes with a Syntax Error. Is this a known problem? Is there a solution?
[18:01:40] <alkar> @Number6: one really final thing if you've got the time, is that inconsistency in oplog gb/hour meter in the new dashboard I made anything to worry about? It looks strange
[18:05:27] <Number6> alkar: That does look strange. Can you do a db.collection.count() on each member of the RS? If that server is off, do an initial sync
[18:25:47] <synesp> our application is in need of a way to track changes in collections - if I field is changed in a document, we need to know what fields and record the old values
[18:26:24] <synesp> I read that one company had great success by running two separate dbs one for their latest app data and another for historical data
[18:28:52] <Nodex> http://www.nodex.co.uk/article/26-05-13/document-record-versioning <--- little background
[20:57:30] <starsinmypockets> Bit of a general / strategic question: I'm coming from sql, where I'd leverage joins with fairly large queries to build up data objects with a lot of fields with associated. Now, working in node with mongoose and mongo I find myself doing a lot of sequential or parallel queries to get the same data. Am I doing it wrong? Stuff like: get the doc, get child elements, get the author, etc etc, then return selected json...
[21:05:07] <cheeser> but if you're always grabbing certain external documents whenever you grab a document, you might consider embedding those docs
[21:05:32] <cheeser> there's a trade off, though. if A "has" a billion "B" documents, embedding is a bad idea
[21:13:36] <starsinmypockets> For instance, I have a set of categories containing embedded subcategories... I need to "loop" through the subcategories and get all associated documents... in sql I would just use joins - in node I'm considering async.each which seems inefficient...
[23:20:10] <edwardbc> I'm trying this: db.users.update({}, {$pull: { contacts:{ $size:0 } } }) but won't work
[23:28:12] <edwardbc> same way this doesn't work db.users.update({}, {$pull: {contacts: {$not: {$size: 0}} } })
[23:28:18] <cheeser> well, you'll probably want: db.users.update({}, {$set: { contacts:[] } }, {multi: true})